SWIR LEDs in Machine Vision: Application Deep Dive and System Design Notes

Short-wave infrared (SWIR) light sources are transforming industrial machine vision. While visible imaging sees color and NIR imaging sees limited material contrast, SWIR illumination sees the chemical and physical signatures of materials – moisture content, polymer type, foreign matter, and subsurface defects. For system integrators and end users in food processing, recycling, semiconductor manufacturing, and pharmaceutical inspection, understanding how to apply SWIR LED light sources correctly decides whether a vision system delivers reliable reject rates or a pile of false positives.

This article is a technical companion to our SWIR buyer’s guide. It focuses on real applications, system-level design choices, and the engineering levers that make SWIR illumination work in production environments.

The physics that makes SWIR useful

Materials interact with light differently depending on wavelength, and the SWIR band – roughly 970–1700 nm – contains several important spectral signatures:

  • Water absorption peaks near 1450 nm and 1930 nm make moisture measurement highly sensitive in the SWIR region.
  • Polymer fingerprints between 1150 nm and 1700 nm allow plastic identification without contacting the material.
  • Silicon transparency above approximately 1100 nm lets SWIR light pass through silicon wafers, revealing subsurface structures.
  • Organic material contrast – fruits, grains, and nuts show maturity, bruising, and foreign contamination in SWIR that RGB cameras cannot see.

This is why SWIR LED chips are not merely “infrared LEDs with a different wavelength”. They unlock spectral information, and the illumination source must deliver that wavelength accurately and consistently under industrial conditions.

Application 1 – Food and agricultural sorting

Modern sorters process 10–15 tonnes of product per hour per channel group. They detect bruises, moisture variance, foreign material (stones, plastic fragments, stems), and internal defects that are invisible in visible light.

Design considerations for SWIR illumination in food sorting:

  1. Wavelength selection – fruit moisture correlates strongly with 1450 nm absorption; foreign bodies like wood and plastic are discriminated in the 1100–1700 nm range.
  2. Illumination uniformity – line-scan cameras need a flat, uniform light line across the full belt width. Uneven illumination creates false “defects” at the edges of the field.
  3. Strobing and duty cycle – high-speed sorters may require pulsed operation. Ask the LED supplier for the maximum pulse current and the duty-cycle limits before designing the driver.
  4. Temperature stability – a sorting hall in summer can reach 40 °C+. Wavelength drift with junction temperature must stay inside the detector filter band. Ceramic-packaged SWIR LED chips with lower thermal resistance are strongly preferred.

Application 2 – Plastic recycling and polymer identification

Recycling lines must separate PET, PVC, PP, PE, and PS fractions that look identical to the eye. SWIR spectroscopy at 1650 nm distinguishes PET from PVC reliably; near 1200–1400 nm, other polymers can be discriminated.

Practical lessons from field installations:

  • Belt speed and material density determine required irradiance. Compute the optical power at the sensor rather than the LED vendor’s “chip output” number.
  • Black plastics absorb strongly across the visible range but remain identifiable in SWIR – a major advantage for mixed waste streams.
  • Dust and contaminant films on the illumination window attenuate SWIR strongly. Design the optical window for easy cleaning, and ask for durable, sealed modules rather than bare chips exposed to debris.
  • Leakage (scattered light) from one channel into another creates false rejects. Barrier optics and controlled beam angles are part of the system design, not an afterthought.

Application 3 – Semiconductor wafer and die inspection

Silicon becomes transparent to light above about 1100 nm. SWIR illumination lets inspectors see through wafers and packaged devices to check die placement, bond integrity, voiding, and hidden cracks.

System design notes:

  1. Wavelength choice trades penetration depth against resolution. Around 1200–1300 nm, penetration is good while resolution remains acceptable for many inspection tasks.
  2. Coherent (laser) versus LED illumination: SWIR LEDs are eye-safer and cheaper for continuous flood illumination; lasers still win where extreme power density or coherence is needed.
  3. Thermal and ESD control in the cleanroom – SWIR devices are ESD-sensitive. Design the driver and handling per the component’s ESD classification.
  4. Wafer thickness variation means the optical path length changes; a stable, narrow-band SWIR source improves repeatability of transmission measurements.

Application 4 – Pharmaceutical inspection and medical diagnostics

Pharmaceutical applications – tablet coating inspection, blister pack integrity, moisture measurements – benefit from the same SWIR signatures. Medical devices use SWIR illumination for hydration sensing and certain diagnostic instruments.

For these markets, traceability and certification matter more than in industrial sorting:

  • Choose components with documented quality systems (ISO 9001) and, where relevant, AEC-Q102 or equivalent stress qualification.
  • Stable wavelength over temperature is critical for calibration-based instruments.
  • Consider lot-level data – if you must reorder the same spectrum months later, binning consistency between lots becomes a contractual issue.

Application 5 – Moisture and liquid level sensing

The 1450 nm water absorption band enables moisture sensors in grain, paper, textiles, and soil; 1930 nm extends sensitivity for low-moisture ranges. SWIR LED-based sensors replace bulky halogen solutions with compact, solid-state sources.

Design lever: the measurement is differential – two wavelengths, one at the absorption peak and one off-peak, compensate for distance and geometry. The source must hold both wavelengths stable, which again points to ceramic packages and documented derating.

System design checklist for SWIR illumination

Before you commit to a SWIR LED module, work through this checklist with your supplier:

  • Required irradiance at the object plane (not at the LED die) – specify in mW/cm².
  • Uniformity target across the field – ±10% is a common industrial target.
  • Operating temperature range and ambient conditions (dust, humidity, washdown).
  • Pulse versus CW operation, duty cycle, and pulse current limits.
  • Optical interface – free space, lens, or fibre-coupled; numerical aperture constraints.
  • Thermal management – heatsink area, airflow, or liquid cooling available in the enclosure.
  • Detector/spectrometer band – keep the LED emission inside the filter acceptance window across the full operating range.
  • Certification requirements – CE, UL, AEC-Q102, RoHS/REACH.
  • Lifetime and maintenance schedule – L70 hours at your actual thermal condition.

Common mistakes in SWIR system design

  • Buying by chip type instead of by application requirement. A “1550 nm LED” without an irradiance specification is not a design input.
  • Ignoring wavelength drift. The detector filter and the LED both shift with temperature; the operating window must be verified at the extremes.
  • Undersizing the thermal design. SWIR LEDs convert less than 10–20% of input power to light at long wavelengths. The rest is heat that must be removed from a small area.
  • Assuming uniformity. Bare-chip arrays need homogenising optics or lensed modules; edge fall-off triggers false rejects.
  • Skipping the pilot test. Sorting material is a moving target – validation should run on production material, at production line speed, before the capital commitment.

Where QUEENDOM fits in your SWIR system

QUEENDOM supplies SWIR LED chips and modules across the 970–1700 nm band, built on ceramic substrates for harsh industrial environments. We publish thermal derating curves, support pulsed applications with documented limits, and can ship evaluation modules with matched lens/homogeniser options for sorting and inspection integrators. If you are designing a SWIR vision system, tell us your object-plane irradiance target and your temperature range – the right starting point for a system that will actually perform in the field.

SWIR is a powerful but demanding illumination technology. The systems that work are the ones designed around thermal reality, wavelength stability, and field conditions – not around the numbers on a datasheet. Use the checklist above, validate with production material, and choose a supplier who treats your application as a system problem rather than a chip sale.

A worked example: sizing the light for a sorting line

Assume a line-scan camera inspects a 1.2 m-wide belt at 3 m/s, and the integration time per line is 60 μs. The detector requires 4 mW/cm² at the object plane to reach the target signal-to-noise ratio. The illuminated area for one exposure is the width of the line (1 cm) multiplied by the belt travel during integration (approximately 0.02 cm), so the instantaneous illuminated surface is 0.024 cm² and the required optical energy per pulse is about 96 μJ. At a line rate of 50 kHz the average optical power needed is roughly 4.8 W at the object plane, before accounting for lens losses (typically 20–30%) and uniformity margins. The practical system will therefore specify an LED module delivering 8–12 W of optical power at the window, with homogenising optics, to respect a ±10% uniformity budget.

This example is simplified – real designs add the detector NEP, the spectral matching factor of the filter, and the duty-cycle limits of the LED – but it shows the correct method: work backwards from detector sensitivity and line geometry to optical power at the object plane, then to module rating with the application’s margins included.

Acceptance testing before production sign-off

Whatever the supplier promises, verify with a structured acceptance test at your facility:

  1. Measure irradiance at five points across the field (centre, two edges, two quarter positions) at the real working distance, at nominal and extreme temperatures.
  2. Record wavelength with a spectrometer before and after a 1-hour thermal soak; compute the drift against the detector filter window.
  3. Run a pulse endurance test at your actual duty cycle for 8 hours; log optical power at the start, midpoint, and end, and record case temperature.
  4. Test with production-representative material, not a clean laboratory sample – dust, vibration, and stray light change results.
  5. Document everything in a test report with photos, settings, and raw logged data; this report becomes the baseline for future supplier audits.

An acceptance test takes days and costs little compared with a production line that discovers false rejects after installation. Build it into every SWIR illumination purchase, and update the baseline at each reorder to catch hidden process drift in the supplier’s manufacturing.

Keeping the terminology aligned with the inspection platform

Specification documents for machine-vision illumination should mirror the naming used across the rest of your lighting portfolio. For the SWIR side, standard terms are SWIR LED Chips in the 970-1700 nm band, frequently supplied in rugged Ceramic LED Chips packages that tolerate the vibration and thermal cycling of a production floor; for automotive inspection cells the same components are specified with AEC-Q102 qualification. When your procurement team compares quotes, an RFQ written with these exact terms returns comparable bins, consistent thermal data, and test reports that can be audited side by side.

Related Products

Further reading: SWIR LED Chips (970-1700nm) · Machine Vision LEDs

Leave a Reply

This site uses cookies to offer you a better browsing experience. By browsing this website, you agree to our use of cookies.
Privacy Policy | Terms of Service | Cookie Policy